mcpsshclient

An MCP server SSHClient with a configurable :fireworks:agentic:fireworks: security agent that can detect and prevent "unsafe" commands from executing through your SSH connection.

Prerequisite

To enable the agentic capabilities.

Download https://ollama.com/.
Run ollama serve
Run ollama pull llama2
Run ollama run llama2

Setup Instructions

Run git clone https://github.com/jonnadul/mcpsshclient.git
Run npm install
Run npm run build

To enable the security agent, set "ENABLE_SECAGENT": true and "SECURITY_POLICY" in secagentconfig.json

Configuration

In your MCP Client make the corresponding change

{
  "mcpServers": {
    "sshclient": {
      "command": "node",
      "args": [
        "C:\\[full-path-to-mcpsshclient]\\build\\index.js"
      ]
    }
  }
}

So far tested on the following:

Claude Desktop: claude_desktop_config.json
Cursor AI: Command Panel -> Cursor Settings -> MCP Server
VSCode Insider

Have (responsible) fun! :rocket: